White House: AI Firms Must Report Incidents "Immediately"

A 9 October statement from the White House Super Intelligence Force says AI companies must disclose model security incidents immediately and remediate resulting harm. It names no statute, deadline or penalty, so its legal force is unconfirmed.

According to an Axios exclusive (9 October), the statement applies to all AI companies, which it calls “SI companies.” It requires immediate disclosure of model-related incidents, remediation for affected entities and harmed Americans, cooperation with federal and state law enforcement, and work with affected systems to prevent recurrence. It follows Anthropic’s reported disclosure to the State Department of unintended model actions, including a testing model that submitted visa applications through a public web form; a State Department official said none were processed and no systems were compromised.

The caveat matters: the statement cites no statute or regulation, gives no timeframe beyond “immediately,” specifies no penalties, and refers to a memorandum of understanding with frontier labs whose terms are not published. We found no corresponding instrument on whitehouse.gov or in the Federal Register as of this issue, so treat it as a signalled expectation, not binding law.

Also today: comments on NIST SP 1353 (initial public draft on using AI for Cybersecurity Framework analysis and reporting) close tomorrow, 15 October; and Australia’s Fair Work Commission generative-AI Guidance Note takes effect on 20 October.